KR20100084179A - Method for caching content data packages in caching nodes - Google Patents

Abstract

A method for caching content data packages in caching nodes 2 of a network 1 comprising a plurality of nodes 2,8 and a plurality of data lines 5 that extend between adjacent of said nodes 2,8, wherein content data traffic is to be routed on traffic connections between a content data library server 7 and user nodes 8 is proposed. The method is comprising the steps of assigning a popularity value of each content data package to each caching node 2 having at least one user node 8 as an adjacent node, calculating a weighted request probability for each content data package at each caching node 2, preferably by combining each popularity value with a distance from the respective caching node 2 to the caching node to which the respective popularity value is assigned to, deciding which of the weighted request probabilities of the content data packages are fulfilling a predefined condition, and caching the respective content data packages in the caching nodes at which the weighted request probability of the content data package fulfils said predefined condition.

Description

How to cache content data packages {METHOD FOR CACHING CONTENT DATA PACKAGES IN CACHING NODES}

The present invention relates to a method of caching a content data package at a caching node (superpeers) of a network comprising a plurality of nodes and a plurality of data lines extending between adjacent nodes, wherein the content data traffic is content. Passed through a traffic connection between the data library server and the user node (peer).

In a typical Internet Protocol Television (IPTV) system, all content is typically located on a central library server, ie stored as a content data package on the library server or ingested at a centralized ingest point. A Content Distribution Network (CDN) is used to distribute content from a library server to a streaming server located near the network edge. If the user wants to view content not available at the edge streaming server, the request is either served or rejected by the library server.

There are two known distribution models, content everywhere and dynamic redistribution. In the first model, the preselected content, i.e. the content data package from the library server, is distributed in advance to all streaming servers. The content is rotated periodically. In the second model, the content is distributed to all streaming servers in advance. However, the library server may redistribute the content based on the usage statistics. For example, if a number of requests are received for an edge streaming server for which no content is available, the central server pushes the content to the streaming server from which the next request is served.

Another current approach to industrial IPTV streaming is based on peer-to-peer (P2P) distribution networks. When a P2P distribution network is augmented by a caching overlay to improve distribution quality, it is referred to as super P2P. Current P2P networks and current super P2P (SP2P) networks do not provide sufficient quality for real-time media, for example video data transmission.

The best known solution for the conventional IPTV content distribution model uses a real-time content distribution technique from the content library server to the streaming server. This technique is referred to as 'stream through'. According to the above technique, if the content is not available at the streaming server (streamer array), the streamer array requests a real-time content segment from the library server (vault array). The streamer then immediately transfers the content to the set top box (STB) and caches it for subsequent requests served from the cache. Some optimizations are applied, for example, by filling the cache in advance.

Another known CDN implementation is based on an owner overlay distribution network optimized by dynamic path reestablishment for elastic media delivery to the edge streaming server.

The best known solution for super P2P video transmission is based on deploying caching nodes close to the edge location and is not suitable for real time viewing experience. According to this technique, a caching node is used to improve P2P distribution by providing storage accessible for the most popular data in networks with high bandwidth.

The first solution relies on a high quality core network to enable real time distribution between the library and the streaming server for real time video content. Distribution delays should typically not exceed hundreds of milliseconds, which requires high quality resulting in expensive core networks. The core network may be unavailable in some cases. The second drawback is that the cache is filled in response to user demand, which creates a bottleneck for long length content and large demand changes, for example, many new highly popular assets are when the content is cached on all edges. Until the initial denial of service. The volume of content chums means that there is insufficient network bandwidth to distribute all content when loaded. The solution is expensive and the main cost contributor is the cost of the bolt array / streamer using owner software and high quality core network, as described above.

The cost disadvantage of the first solution is solved by the second solution, namely super P2P. SP2P is a relatively inexpensive device that runs on a public Internet network. However, the major disadvantage of current SP2Ps is that they cannot deliver real-time content and are not designed to take into account real-time service requirements. In addition, the volume of content (long length) means that not all content can be stored in every place.

Accordingly, an object of the present invention is a method of caching content data packages at a caching node that solves at least one of the problems associated with the related art, in particular to improve the efficiency of the SP2P network which enables the delivery of live video and real-time services. To provide.

According to the invention, this object is achieved by a method of caching a content data package at a caching node of a network comprising a plurality of nodes and a plurality of data lines extending between adjacent nodes, wherein the content data traffic is cached. Passed through a traffic connection between the library server, caching node, and user node.

That is, content data traffic is routed through a traffic connection between the content data library server and the user node, which is not merely understood in the conventional sense of the term “root” and selects a route in the network to transmit data. The term is associated with physical transmission. In the case of the present invention, the term "root" assumes that a cache layer exists in the network to enhance traffic, and how to populate this cache layer to optimize network usage. Cover it. Push and pull transmission is possible. In pull mode, a cache or peer that requires a piece of content requests it from a caching node.

The method of the present invention includes the following steps. In other words,

Assigning a population value of each content data package, segment or part to each caching node having at least one user node as an adjacent node. Accordingly, different populations in different areas where nodes are located may be considered to define the population values to be assigned. In the case of the method of the invention a broader "package" is used. Typically a package is associated with a small amount of data, for example a UDP packet. This is a small unit that builds population density, whereas larger segments or entire portions can be evaluated and cached for populationity.

Preferably calculate the weighted request probability for each content data package at each caching node by combining the populationity value with the distance from each caching node to the caching node to which each respective population value is assigned. step. The request probability is proportional to the assigned population value.

Determining which of the weight request probabilities of the content data package performs a predefined condition. Thus, caching decisions are made, for example, according to a predefined formula.

Caching by each of the content data packages at the caching node where the weight request probability of the content data package performs the predefined condition.

The method of the present invention improves the efficiency of the SP2P network to enable the delivery of live video and real-time services as content data packages by introducing layers into population clusters and intelligent caching on different 'popularity' layers. Thus, the efficiency of a managed on-demand content delivery network, preferably a super peer to peer network (SP2P) or caching overlay network, is improved by introducing layers into the population clusters and intelligently caching on different layers.

The method of the present invention results in the implementation of two concepts as follows.

The first concept is to introduce hierarchical population layers into media or media segment caching with horizontal mesh interconnection and to place caching nodes in the population cluster layer (overlay network). The next step is to apply the population density to determine which medium (or medium segment) should be cached and in which layer as described below. Typically the one closest to the user caching node with the weighted highest population density should be cached or chosen for caching.

The second concept is to advance the overlay network implied in the first concept by pre-populating the cache based on the population topology topology found during the first concept. In one embodiment provided for illustrative purposes, the population topology for a given genre may be derived from a self-tuning caching topology. The genre population topology can then be used to pre-populate the cache with media from the same genre and high estimated population.

The method of the present invention has the following advantages. In other words,

Improve the efficiency of content delivery on public Internet networks.

Results in an SP2P network optimized for real-time services.

Allows conventional IPTV services such as video on demand, broadcast TV, network PVR, live-pause TV on public Internet networks or SP2P networks.

Eliminate the cost of distributed video servers and high quality core networks from IPTV services.

-Consider media characteristics for efficient content delivery.

Arrange resources as media assets and / or segments, most of which are requested and valued;

Reduce capital and operating costs for next generation IPTV services.

Based on widely available Internet networks.

Allow content owners to reach end users directly.

Provides standardization potential for NGN Release 2.

The method of the present invention is important for cable companies that optimize SP2P and / or open Internet networks for real-time conventional IPTV services and expand into IPTV markets as well as IPTV vendors.

Preferred Modifications of the Invention

A particularly preferred variant of the method of the present invention is that after a predetermined time period, the population value is reassigned by a new population value, the new population value being used for the content data package during the time period. It is characterized in that it is evaluated in proportion to. Thus, reassessment of the initially assigned population value can be achieved based on actual usage data.

Advantageously, said weight request probability of said content data package is calculated by summing the results of said combining for different caching nodes and / or user nodes. Thus, data can be cached at a node without an adjacent peer in an appropriate manner in accordance with the above-listed concepts. Any node in the network can be used as the caching node. Caching nodes are defined by the fact that they are caching the content data package for transmission to at least one of the peers in the network. As described above, the transmission may be done in push or pull mode.

The weight request probability for each content data package is calculated by each caching node and the determination is made by the caching node that calculated the request probability to enable decentralized management of the network.

Advantageously, said weight request probability for a content data package, ie a weight request probability threshold for caching media, is requested and / or updated from an external source, preferably from the data library server (library node). Thus, when new high population content is added, less popular media can be pushed out of the cache.

The method of the present invention is preferably used to cache a content data package containing live video data and / or real time service data.

In a preferred variant of the method of the present invention the weight request probability for the content data package is calculated by a data distribution node and the determination is made by the data distribution node and the respective content data package, i.e. a new multimedia asset. Before being routed and / or transmitted for a user node that has requested at least one of each of the respective content data packages, the content data package is configured for the caching node in which the weight request probability of the content data package performs the predefined condition. Is pushed against. Thus, precaching of content data packages is possible.

In the latter case, the data distribution node includes the content data library server. Thus, the content data library server can act as a central node for managing the network.

Advantageously, upon request for a content data package cached at at least two of the caching nodes to be transmitted to a user node of the user nodes, the requested content data package is from both caching nodes in parallel or sequentially. Sent in push or pull mode for the user node. Thus, a service request for content data packages from multiple sources is possible, which is useful when there are two caching nodes each serving partially overlapping with a group of cache users. Non-overlapping portions of the group may produce peak loads at different peak times. Having media segments, ie content data packages, on different caching nodes allows selection of nodes with a minimum current peak load. Alternatively, when one of the nodes is busy, it can send distributions to other caching nodes and, for example, focus on new requests for content data packages from peers in the network.

Also within the scope of the present invention is a computer program comprising means for performing the method of the present invention when executed and / or stored on a computer system.

The method of the invention can be implemented on a general purpose computer (hardware) as well as integrated into edge equipment such as edge and clustered cache nodes such as DSLAM / ISAM and routers.

Advantages can also be obtained from the description and the disclosed drawings. The features mentioned above and below can be used individually or collectively in accordance with the invention in any combination. The examples mentioned are not to be understood as complete enumerations and have exemplary meanings for the purposes of the present description.

The invention is shown in the drawings,
1 is a diagram illustrating a managed content distribution network having multiple cluster levels for caching configured to perform the method of the present invention;
2 is a diagram illustrating services of several requests from multiple sources according to a preferred embodiment of the present invention.

1 shows a network 1 designed to carry out the method of the invention for caching content data packages at a caching node 2 of the network 3. The network 1 comprises a plurality of nodes and a plurality of data lines 5 extending between adjacent nodes. Content data traffic is routed through a traffic connection between the content data library server 7 and the user node 8, i. E. A client node or a peer. A sample algorithm will be described to illustrate the method of the present invention. The proposed algorithm aims to provide a more general solution than a specific algorithm for population topology generation or caching.

In the drawing of a client node or peer, one example of an STB is connected to edge caching nodes referred to as caching nodes South CN_S, East CN_E, West CN_W, and North CN_N. These caching nodes are in turn connected to caching nodes on higher cluster layers CA_1, CA_2, ... CA_N.

Media assets are initially stored as content data packages in a central media store, ie, a content data library that can be distributed in a single location or in multiple locations.

According to the first concept described above, each super peer distribution node that clusters segments from other nodes when content data is transmitted to a peer has a population value, i.e., for example, at least one user node adjacent to it. Caching content based on the content population level derived from the multiple population levels introduced for each content data package of each content data package assigned to each caching node and the distance to the requested peer as a node. The decision is made to

One of the definitions of "popularity density" is the population per user. For example, movie A was requested 100 times for region A and 100 times for region B. There are 1000 users in the area A and 10000 users in the area B. In this example, despite having recorded the same number of requests, the population density is ten times higher in area A than in area B.

"Popularity level" means that the "popularity density" is subdivided into multiple levels, for example, by different levels of populational density.

For example, the caching decision may be based on whether the population density is high enough relative to a node close to the customer location. In general, the weight request probability for each content data package at each caching node may, for example, vary each population value from each caching node to the caching node to which each population value is assigned. Calculated by combining with distance.

To make a caching decision, it is determined which of the weight request probabilities of the content data package performs a predefined condition, for example represented by the following formula.

The determination of whether to cache the content may be made by the super peer node itself, by a central library server or other agent, or by the super peer node and the library server in combination. For example, the super peer node may request a library server for the population topology of the surrounding node before making a decision.

The introduction of the layer into the population topology can be illustrated by another example. For example, three portions of content A, B, and C are equally popular in the south east region overlaid by caching nodes CN_E, CN_S, CN_A1. Content B is popular in South, C is popular in East, A is less popular in South and East, but is popular throughout the crowded area South East.

In the context of this proposal, the inventors apply broad meanings based on any combination of the following factors, but are not limited to the following list.

Box office or other external population marker

·genre

User profiles who know what and where they prefer

History data

· Categorize your content, for example, if it's available in the 'Your TV' community

Monitor recommendations to estimate where content will be required

·Etc

These factors are examples of the background features that define the weight request probabilities for each content data package, i.e., the popularity for each content data package to be requested by different peers.

In this case, by introducing the caching nodes CN_E, CN_S and CN_A1, the transmission of contents A, B, and C caches B on CN_S, caches C on CN_E, and southeast by caching A at the cluster caching layer CN_1. Can be localized for the area. This optimizes the efficiency of the CN_E and CN_S caches with more cache hit ratios-more requests sent from externally requested caches.

As shown above, the first concept improves the efficiency of SP2P content by introducing multiple levels of population that enable efficient placement of resources such as content and load, which are mostly requested for real-time services. Has value.

The second concept described above is based on the population topology found during the steps of the method of the invention, ie the method according to the method of the first concept described above, for example, for a new multimedia asset. Further advance the first concept by pre-populating the cache. In one embodiment, the population topology map for a given genre is derived from the media population topology. The map is used to pre-populate the cache for new media in the same genre category. For example, if the library server or external agent has the same genre and rating as A, it may decide to move new content, that is, content data package D, to the caching node CN_A1, which is the efficiency of an open network that enables real-time services. To further improve.

One embodiment of an algorithm for caching in the first concept as providing an example of a formula for the caching decision is shown below. If D _n is the distance from the nth cache node to the requested peer, e.g. the number of nodes between them, then R _n is the number of observers or predicted requests for itself at node n, i.e., of the content data package. Request probability proportional to the population value. The asset, ie the content data package, must then be cached at one or more node (s) K with the highest weighted population density, ie, weighted request probability.

Averaging content is weighted probability defined by a representation in parentheses, in the region near the maximum p of all nodes that define a predefined condition performed by the caching node where the content data package is cached in accordance with the method of the present invention. It should be cached at node K with the value of p.

Thus, it is contemplated that several peers with multiple access paths may request the same segment, that is, the content data package. Therefore, a caching node that is optimally located to serve all peers should be selected. This is a simplified way of expressing the equation.

2 illustrates another optimization of the method of the present invention that can be applied by servicing several requests from multiple sources. This increase fills the cache's time but eliminates the load from a single network link that allows it to trade time for link capacity.

Upon request from the STB 20 as a peer for the content data package T1 cached at the two caching nodes CN_A1 and CN_A2 shown, the content data package is sent to the STB as a user node. Each content data package is optionally delivered to the user node from all caching nodes that are in parallel or in sequence across the caching node CN_E having the content data package for T1. All illustrated caching nodes are used as pull interfaces. The content data package T1 is subdivided into segments T1-1 and T1-2 transmitted over the data line 5, for example, using a data rate of 1 Mbps. The content data package T1 comprising segments T1-1 and T1-2 is transmitted from CN_E to the illustrated STB via data line 21 using a data rate of 2 Mbps, for example.

A method of caching content data packages at a caching node (2) of a network (1) comprising a plurality of nodes (2,8) and a plurality of data lines (5) extending between adjacent nodes (2,8). This proposed, content data traffic is delivered via a traffic connection between the content data library server 7 and the user node 8.

This method,

Assigning a population value of each content data package to each caching node 2 having at least one user node 8 as an adjacent node,

Preferably each content data package at each caching node 2 by combining the respective population values with the distance from each caching node 2 to the caching node to which the respective population value is assigned. Calculating a weighted request probability for;

Determining which of the weighted request probabilities of the content data package performs a predefined condition;

Caching the respective content data package at the caching node where the weight request probability of the content data package performs the predefined condition.

The proposed solution introduces multiple levels of population, intelligently caches on different 'popularity' layers, and enables efficient deployment / use of resources that are mostly requested and valuable, thus enabling efficient content delivery on public Internet networks. To improve. This allows the use of public networks overplayed by cost-effective layer caching infrastructure for common IPTV services such as video on demand, broadcast TV, network PVR, live pause TV. Another advantage of the solution is that it does not require a network of video servers and a high quality core network as in the prior art. The proposed invention improves the efficiency of SP2P networks that generally enable the delivery of live video and real-time services.

Claims (10)

A method of caching content data packages in a caching node of a network comprising a plurality of nodes and a plurality of data lines extending between adjacent nodes.
Content data traffic is routed and / or forwarded through a traffic connection between the content data library server and the user node,
The method comprises:
Assigning a populationity value of each content data package to each caching node having at least one user node as an adjacent node,
Calculating weight request probabilities for each content data package at each caching node;
Determining which of the weighted request probabilities of the content data package performs a predefined condition;
Caching the respective content data package at the caching node where the weight request probability of the content data package performs the predefined condition.
How to cache content data packages.
The method of claim 1,
The weight request probability is calculated by combining each populationity value from each caching node to the distance from each caching node to which each populationity value is assigned.
How to cache content data packages.
The method of claim 1,
After a predetermined time interval, the population value is reallocated by a new population value, and the new population value is evaluated in proportion to the frequency of use of the content data package during the time period.
How to cache content data packages.
The method of claim 1,
The weight request probability of the content data package is calculated by summing the results of the combining for different caching nodes and / or user nodes.
How to cache content data packages.
The method of claim 1,
The weight request probability for each content data package is calculated by each caching node and the determination is made by the caching node that calculated the request probability.
How to cache content data packages.
The method of claim 1,
The weight request probability for a content data package is requested and / or updated from an external source, preferably from the data library server.
How to cache content data packages.
The method of claim 1,
The content data package includes live video data and / or real time service data.
How to cache content data packages.
The method of claim 1,
The weight request probability for the content data package is calculated by a data distribution node and the determination is made by the data distribution node, the user node for which each content data package requested at least one of the respective content data packages. The content data package is routed to and / or forwarded to the caching node where the weighted request probability of the content data package performs the predefined condition prior to being routed to and / or sent to.
How to cache content data packages.
The method of claim 8,
The data distribution node includes the content data library server.
How to cache content data packages.
The method of claim 1,
Upon request for a cached content data package at at least two of the caching nodes to be routed and / or delivered to one of the user nodes, the requested content data packages are parallel or sequentially Routed and / or forwarded from both caching nodes to the user node
How to cache content data packages.

Images